Abstract 40: Updating White Matter Hyperintensity Visual Scoring: Incorporating Research Advances, Improving Reproducibility and Correlation with Automated Volumes
Introduction: White Matter Hyperintensities (WMH) have been implicated as a risk factor for motor and cognitive decline, dementia and stroke. A standard for use in clinical settings is needed to identify which cases are advanced or meet thresholds with implications for degree of impairment. The system of Fazekas is a promising candidate due to its simplicity and strong correlation with outcomes. The original Fazekas categorized deep and periventricular WMH separately. Advanced periventricular and deep WMH were later shown to have equivalent pathology and etiology. Subsequent work by De Carli showed the classification of deep and periventricular WMH on axial images to be arbitrary and suggested a common etiology for lesser degrees of WMH as well. We therefore adapted the Fazekas criteria to consider deep and periventricular WMH jointly. Questions persist about low reproducibility of grading systems. In a preliminary analysis we found frequent disagreement classifying lesions less than 3mm in diameter and decided to incorporate this threshold into our criteria to determine the effect on reproducibility and correlation of our final model with automated volumes.
Hypothesis: Modification of the Fazekas grading system to consider deep and periventricular WMH jointly is less arbitrary. Simple modifications of such a system will result in a high degree of reproducibility and high correlation with automated WMH volumes.
Methods: Axial 3T FLAIR MR images of the brain were obtained from community dwelling subjects. Grading measurements were applied to diameter of deep WMH and thickness of periventricular WMH. The initial system defined Grade 0 as no WMH, Grade 1 as < 10mm, Grade 2 as >=10mm but <20mm and Grade 3 as >=20mm. 52 studies were read separately by 2 reviewers. We then revised grade 0 to include intensities < 3mm and grade 1 as >=3mm but <10mm. 40 additional studies were then read and level of agreement was re-assessed. 563 studies were read for correlation with automated volumes.
Results: After modifications, inter-rater agreement for Fazekas white matter score increased from 72% to 89% with kappa increased from 0.45 to 0.78. There was good correlation of grade and automated volume (ANOVA R-Square: 0.52 P <.0001).
Conclusion: Advances in understanding of White Matter Hyperintensities suggest periventricular and deep lesions should be rated jointly. This modification was applied to the Fazekas system with excellent reproducibility after implementation of a 3mm size threshold as well as high agreement with automated volumes ammong community dwelling subjects.
- © 2012 by American Heart Association, Inc.